New UW, PNNL Institute Attracts Supercomputing Expert Dunning
The Northwest Institute for Advanced Computing (NIAC) has landed supercomputing luminary Thom Dunning Jr., who will help lead the effort to tie together two of the region’s top centers of computing research.
Dunning comes to the NIAC after a career that has included stints at several universities and national laboratories—including Pacific Northwest National Laboratory, which is partnering with the University of Washington to form the new institute, first announced in January. Most recently, Dunning led the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign, where he oversaw completion this spring of the $300 million Blue Waters supercomputer.
“We have only begun to tap the power of computing to advance our knowledge of the world around us and to improve the lives of our citizens,” Dunning says in an e-mail. “By bringing UW and PNNL together in the Northwest Institute for Advanced Computing, we will be able to capitalize on the combined strength of these two institutions to take full advantage of these many opportunities.”
Dunning is taking over as co-director from PNNL’s Moe Khaleel. UW electrical engineering chair Vikram Jandhyala is the other co-director.
The NIAC, situated in Sieg Hall on the UW campus in Seattle, is designed as a center for collaboration among researchers from both institutions—previously separated by a three-plus-hour drive—focusing on new technologies to advance computing, data-enabled discovery, and computational science.
“It’s amazing how just a few meetings in person makes a big difference,” Jandhyala says.
While still in its early stages, the institute has the potential to complement many of the region’s existing strengths—in enterprise IT, cloud computing, machine learning, mobile, and high-performance computing—that position it for leadership in big data, the broad area of innovation around capturing, managing, and analyzing data in ever-expanding quantities and forms.
Regions around the country are vying to build big-data clusters with efforts across academia, government, and industry. These include Boston, New York, and surrounding areas, where financial services and marketing firms are adopting these technologies; Silicon Valley, which is a wellspring of big data and analytics companies; and Washington, DC, where national security and intelligence work is increasingly data-driven.
The leaders of the NIAC say the complementary interests and personnel of the two institutions will open opportunities otherwise unavailable. Richland, WA-based PNNL, operated by the Battelle Memorial Institute, focuses more on applied research, typically oriented toward problems faced by the Department of Energy and national security agencies. UW has a broader research portfolio, an established program in data-driven scientific discovery, an educational mission, and a location in the heart of the Seattle technology cluster.
“You can imagine that together we can successfully compete for research projects that neither of us could do independently,” says UW computer science professor Ed Lazowska, who has been a key liaison to PNNL through his service on an advisory committee to its Fundamental and Computational Sciences Directorate.
The NIAC is also being structured as an engine for new-company formation. An industry consortium and advisory board will include companies and investors interested in the technologies emerging from NIAC.
“It’s going to take a little bit of time, but the idea is to generate interest in funding startups,” says Jandhyala, who is also a UW Presidential Entrepreneurial Faculty Fellow and founder of Nimbic, which makes cloud-based tools for electronics designers. “And there is interest. [UW] President Young wants to double the number of startups coming out, and I think we’ll be able to leverage that energy in the university as well.”
Policy changes at the UW around technology licensing and company formation have made setting up the NIAC with this sort of entrepreneurial focus easier than it would have been in the past, he says. One novel arrangement is the creation of “distinguished faculty fellow” positions, which will allow PNNL researchers working under the NIAC to “have a foot in UW,” Jandhyala says, an arrangement to be reciprocated for UW researchers with PNNL.
“Traditionally, universities have not figured out how to have a faculty person here who also has a foot outside, without IP issues and so on,” he says. “I think that’s been worked out quite well in this case.”
The NIAC will focus broadly on how advanced computing systems can be used in scalable design, modeling, and simulation, as well as data-driven discovery applied to a wide range of fields.
“Ultimately, this will tie in to the science of urban and metro areas, looking at optimizing entire systems where human beings play a key role, including the power grid, cybersecurity, traffic and transportation systems, communication, development, and the environment,” Jandhyala says.
The NIAC provides a conduit to “real-world problems”—critical to training data scientists—being tackled by PNNL in these and other areas, Jandhyala says. “Here’s where the connection to PNNL becomes very important, because there’s datasets which nobody else can have access to,” he says.
The lab is participating in some of the most data-intensive projects in the world right now. For example, it is preparing to host data from the Belle II high-energy particle physics experiments, which will be performed in Japan, beginning in 2015.
PNNL’s relationship with the Belle project, hosted at KEK in Japan, deepened after the 2011 Fukushima earthquake and tsunami, when PNNL stepped in to help the international community of scientists carry on with data analysis of the Belle experiment, says Dick Russell, manager of high-performance computing in the energy cluster of PNNL’s Computational Science and Mathematics Division.
Some 240 petabytes of raw data are expected to be stored at PNNL, more than the projected output of the Large Hadron Collider at CERN. “Right now we think it’s one of the biggest envisaged datasets,” Russell says.
Work is going on now at PNNL to help design the experimental apparatus and plan for transferring a full backup copy of the data from Japan via undersea cables.
Meanwhile, both UW and PNNL have been national leaders in advancing data-driven discovery. Coordinated work in this area will be a key focus for NIAC.
“Most fields of discovery are transitioning from data-poor to data-rich,” says Lazowska, who leads the eScience Institute, which is the center of this work at UW. “The world is full of tiny but powerful sensors—in telescopes, in gene sequencers, in roads and bridges and buildings, in our environment, in the form of Twitter feeds and Web requests. The challenge today is converting all of this data into knowledge, and converting this knowledge into action.”
The goal of the eScience Institute is to make UW a leader in inventing new approaches to data-driven discovery, and also in making these new approaches usable by researchers in a broad range of fields.
Dunning, the new NIAC co-director, also has deep expertise in this area, having helped create the Department of Energy’s Scientific Discovery through Advanced Computing program.
That’s one of several assets he brings to the NIAC, and in turn to the Northwest’s growing big-data cluster.